Send us a text 🌍 INAI • The Open AI Hub The Intelligence Atlas → the world’s most comprehensive, open hub of AI knowledge. 2 Million+ tools, models, agents, tutorials & daily news —free for all, updated every day. https://github.com/inai-sandy/inAI-wiki Top Highlights: Google Gemini 3 and Anthropic Claude Opus 4.5 launch major upgrades with price cuts, intensifying AI model competition. OpenAI faces legal pressure and $200B cost forecasts while developing new devices and platform features. Booking.com and DHL deploy production AI agents at scale; Microsoft releases Fara-7B for on-device automation. China leads US in open model downloads, highlighting community ecosystem strength. Research advances include evolution strategies for billion-parameter transformers, oncology benchmarks, and AI aftershock prediction. New Tools: FLUX.2 offers 4MP image generation with open weights. Tencent's HunyuanOCR (1B) provides state-of-the-art accuracy at lower cost. Hunyuan 3D generates 3D assets from text in minutes. Pinokio 5.0 enables local model serving. dnet allows distributed inference on Apple Silicon. Retake introduces post-render AI video editing. LLM Updates: Claude Opus 4.5 improves reasoning, coding, and browsing while leading benchmarks. Gemini 3 integrates into Search with competitive results and price cuts. Microsoft's Fara-7B delivers on-device Windows automation. DR Tulu-8B outperforms larger models on HealthBench. Grok-4 achieves top Mensa Norway score. Research: NVIDIA/Oxford revive evolution strategies for transformers. CMU identifies LLM-RL bottlenecks. New theory explores context-parameter equivalences. MTBBench introduces oncology benchmark. LLM judge detects ASR errors at clinician-level accuracy. AI predicts aftershocks in seconds. Industry: OpenAI faces court-ordered disclosure and mental-health lawsuits. HSBC estimates $200B needed by 2030. Google-Broadcom deepen TPU partnership; sfcompute raises $40M. Italy probes WhatsApp AI; Georgia launches AI literacy programs. Tutorials: Redis/DeepLearning.AI course on semantic caches. Baseten explains LLM latency bottlenecks. LangChain details agent testing. TPU explainer clarifies VLIW architecture. Demos: Sparse autoencoders steer LLM behavior. FLUX.2 quantization demo. DeepMind AlphaFold documentary. GPT-5.1 builds iOS app in 2 hours. Discussions: Open ecosystems reshape power dynamics. RAG anchors production workflows. Multi-agent systems risk token overspending. AI-native IDEs emerging. Sutskever emphasizes research over compute. Support the show 🌍 INAI • The Open AI Hub The Intelligence Atlas → the world’s most comprehensive, open hub of AI knowledge. 2 Million+ tools, models, agents, tutorials & daily news —free for all, updated every day. https://github.com/inai-sandy/inAI-wiki
